To achieve non-invasive blood glucose in vivo detection, solving common problems such as low measurement sensitivity, large interference of excitation source fluctuation, and serious patient temperature interference is first necessary. In this study, a new method based on temperature compensation for differential photoacoustic measurement was proposed, which can greatly suppress the influence of temperature. Two identical photoacoustic cells were used, one for accommodating the blood to be measured, the other for pure water. Measured and reference photoacoustic signals are respectively obtained. First, the temperature coefficient of the two photoacoustic signals is determined by changing the liquid temperature of the two photoacoustic cells. Then, the error of temperature in the two photoacoustic signals are separately compensated and corrected to eliminate the interference of temperature on the experimental results. Finally, the intensity of one photoacoustic signal is divided by that of the other photoacoustic signal to suppress the influence of laser intensity fluctuation. Based on this procedures the study used tissue engineered skin to simulate a human skin environment and explored the skins permeability to laser and photoacoustic signals. The results show that the R2 of linear fitting between glucose photoacoustic signal intensity and glucose concentration under a low concentration was 0.970 6, thereby proving the feasibility of this method for use in practical applications. The results also show that the average transmittance of the simulated skin to laser was 53.88%, and the average transmittance of the simulated skin to the photoacoustic signal was 9450%. The results of this study can provide a useful reference for the noninvasive detection of blood glucose.